The headline you keep reading, that AI is killing Google, does not match what people actually do. The 2026 data on Google and AI search tells a quieter, more useful story: buyers are not choosing one over the other. They use both, often within the same task. Similarweb found that 95% of ChatGPT users also show up in Google’s audience, and Google’s household reach held steady near 49% even as ChatGPT climbed from 4.6% to 7.9%. Almost nobody left. They simply added a new habit on top of an old one.
That distinction matters more than any scary headline, because it changes what you should do next. If users had abandoned Google, you would rebuild your whole strategy. Since they layer AI on top of Google, you need to win visibility in both places and understand which tasks move where. Here is what the numbers really show, and how to keep your traffic in a two-engine world.
Are people really leaving Google for AI search?
No, and the overlap numbers make that clear. Between March and May 2026, Similarweb counted 461 million of ChatGPT’s 494 million users, roughly 95%, also using Google in the same window. That share sat at 95% in September 2025 and stayed at 95% in May 2026. Use of Google has not declined. More than half of people now start a search by opening an AI app, but that figure is not rising year over year either.
So the loyalty story is settled: people keep Google and add AI. The behavior story, though, is where the real shift hides. Adding AI to the mix quietly changes how many searches people run, what they search for, and how often they click through to your site.
What the overlap numbers actually hide
Stable audience size masks a real drop in activity. A Bocconi University study found that traditional search queries fell 9.4% among households that gained ChatGPT Search access, and that decline deepened to 17.0% after 20 weeks. The people stayed. Some of their searches did not.
The pain lands unevenly by content type. Referral losses concentrate where AI answers the question outright.
| Content category | Referral change after AI access |
|---|---|
| Academic sites | -32.8% |
| Reference sites | -26.5% |
| Developer sites | -15.1% |
| News sites | -13.4% |
| Marketplace and entertainment | Minimal change |
Click behavior inside Google’s own AI Mode shifts too. A field experiment by Wang and colleagues recorded an 18.8 percentage point drop in clicks to external sites under AI Mode, with news clicks down 12.5 points, Reddit down 21.2 points, and Wikipedia down 9.9 points. About 15.3% of AI Mode users said they had trouble reaching a specific website. The audience is loyal to Google, but the click you used to earn is harder to win. This is the same trend behind the measured decline in traditional search.
Which tasks go to AI and which stay on Google?
People sort their tasks between the two engines with surprising consistency. Research, advice, and writing drift toward AI, while navigation, shopping, and quick local lookups stay on Google. An Orbit Media survey put numbers on the split.
- Tasks moving to AI: step-by-step instructions (40%), medical information (36%), vacation planning (35%), recommendations (31% to 32%), and quick factual lookups (29%, up from 23%).
- Tasks staying on Google: local and business queries, where only 24% prefer AI because AI tools still lack strong map integration.
Two more numbers frame the moment. Around 72% of people use AI at least daily, yet most of those prompts serve non-search purposes. And 70% now see Google’s AI Overviews, which appear in about 76% of search results. AI is everywhere in the journey, but Google still owns the local and transactional end of it, which keeps local SEO and the map pack resilient.
Why audience size no longer equals traffic
Here is the trap for marketers. You can watch your Google audience hold steady and still lose visits, because AI now answers many questions before the click. A stable user base does not promise stable referral traffic. Informational content feels this first, since an AI summary can satisfy the reader without sending anyone to your page.
Attribution gets murkier too. A branded search that looks organic may have started as an AI recommendation, which hides the real source of the visit. So the old habit of judging health by sessions alone breaks down. You need to track queries, sessions, clicks, and trust as separate signals, because they now move independently. That is exactly the shift behind reporting AI visibility metrics instead of organic traffic alone.
How to win in a Google and AI search world
You do not need to pick a side. You need to earn visibility across both engines and cover the tasks each one owns. Four moves matter most.
Be the source both engines cite
AI tools summarize retrieved content, so they need a clear source to pull from. Write specific, verifiable, well-structured pages that an engine can quote cleanly. Reassuringly, 85% of buyers still verify AI answers with a follow-up search, which means your page can win the click at the verification stage even after the AI names you. Sharpen your pages using the same principles behind getting cited by AI search.
Protect and differentiate your informational content
Generic answers now get commoditized by AI summaries. Original research, first-hand experience, data, and strong opinions survive because an AI cannot generate them from thin air. If your article only restates what everyone already knows, expect the summary to replace your click. If it adds something new, expect the summary to send readers your way to verify it.
Lean into local and transactional strength
Google still wins local and business queries, and only 24% of people prefer AI for them. Keep your Google Business Profile sharp, earn reviews, and build location pages that answer real buyer questions. This is the part of the journey AI has not taken, so defend it hard.
Track the metrics that move separately
Stop judging success by one blended traffic number. Watch query volume, session counts, click-through rate, and brand mentions inside AI answers as four distinct trends. When one falls while another holds, you learn where AI is intercepting demand and where you still win the visit.
Key takeaways
- People are not leaving Google for AI search. Around 95% of ChatGPT users also use Google, and Google’s reach has stayed near 49%.
- Stable audience size hides a real activity drop: traditional queries fell 9.4% and reached 17.0% after 20 weeks of ChatGPT access.
- Referral losses hit academic, reference, and news content hardest, while marketplace and local searches barely moved.
- Google’s AI Mode cut external clicks by 18.8 points, so earning the click is harder even when the audience stays.
- Tasks split predictably: research and advice go to AI, while local, shopping, and quick lookups stay on Google.
- Track queries, sessions, clicks, and trust separately, because audience size no longer predicts traffic.
Frequently asked questions
Are people abandoning Google for ChatGPT?
No. About 95% of ChatGPT users also use Google, and that overlap held steady through 2026. Google’s household reach stayed near 49% while ChatGPT grew, which shows people add AI rather than replace search.
If users still use Google, why is my traffic falling?
Because audience size and traffic are not the same. AI now answers many questions before the click, so query volume and referral traffic can drop even while your Google audience holds steady. Informational content feels this first.
Which searches do people take to AI instead of Google?
Research, step-by-step instructions, medical questions, trip planning, and recommendations shift toward AI. Local searches, shopping, navigation, and quick business lookups stay on Google, partly because AI tools still lack strong maps.
Does SEO still matter with AI search?
Yes. AI tools summarize retrieved content, so they need well-structured sources to cite, and 85% of buyers still verify AI answers with a search. Strong, specific content wins both the citation and the verification click.
How should I measure success across Google and AI search?
Track query volume, session counts, click-through rate, and brand mentions in AI answers as separate signals. They now move independently, so a single blended traffic number hides where AI is intercepting your demand.
The smart response to all this is not panic, and it is not ignoring AI either. People have shown you exactly what they want: Google for local, shopping, and quick answers, AI for research, advice, and planning, and often both for a single decision. Meet them in both places. Write pages worth citing, defend your local and transactional turf, and measure the signals that actually move. Do that, and the shift from one engine to two becomes an opportunity instead of a threat.
